On 30 Nov 2001, H. Peter Anvin wrote:
> Followup to: <Pine.LNX.4.40.0111301614000.1600-100000@blue1.dev.mcafeelabs.com>
> By author: Davide Libenzi <davidel@xmailserver.org>
> In newsgroup: linux.dev.kernel
> >
> > The point is why store kernel pointers in global registers when You can
> > achieve the same functionality, with a smaller patch, that does not need
> > to be recoded for each CPU, without using global registers.
> >
>
> Because global registers are faster! This is exactly the kind of
> stuff that is properly CPU-dependent and should be treated as such.
> Heck, it even depends on what kind of multiprocessor architecture, if
> any, you're using!
>
> That being said, I belive that on most, if not all, processors, the
> idea of having the pointer point not to "current" but to a per-CPU
> memory area is *very* appealing, and a change that should be made
> uniform unless it's a significant lose on some machines...
Again this is the "current" diff :
static inline struct task_struct * get_current(void)
{
- struct task_struct *current;
- __asm__("andl %%esp,%0; ":"=r" (current) : "0" (~8191UL));
- return current;
+ unsigned long *tskptr;
+ __asm__("andl %%esp,%0; ":"=r" (tskptr) : "0" (~8191UL));
+ return (struct task_struct *) *tskptr;
}
that will probably resolve in something like:
movl %esp, %eax
andl $-8192, %eax
movl (%eax), %eax
- Davide
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Fri Nov 30 2001 - 21:00:41 EST